Npu grpo fix2

.dev_scripts/build_docs.sh

@@ -0,0 +1,7 @@
+cd docs
+rm -rf build
+
+# update api rst
+#rm -rf source/api/
+#sphinx-apidoc --module-first -o source/api/ ../modelscope/
+make html

.dev_scripts/ci_container_test.sh

@@ -0,0 +1,40 @@
+if [ "$MODELSCOPE_SDK_DEBUG" == "True" ]; then
+    # pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
+    git config --global --add safe.directory /twinkle
+    git config --global user.email tmp
+    git config --global user.name tmp.com
+
+    # linter test
+    # use internal project for pre-commit due to the network problem
+    if [ `git remote -v | grep alibaba  | wc -l` -gt 1 ]; then
+        pre-commit run -c .pre-commit-config_local.yaml --all-files
+        if [ $? -ne 0 ]; then
+            echo "linter test failed, please run 'pre-commit run --all-files' to check"
+            echo "From the repository folder"
+            echo "Run 'pre-commit install' install pre-commit hooks."
+            echo "Finally run linter with command: 'pre-commit run --all-files' to check."
+            echo "Ensure there is no failure!!!!!!!!"
+            exit -1
+        fi
+    fi
+
+    pip install decord einops -U -i https://mirrors.aliyun.com/pypi/simple/
+    pip uninstall autoawq -y
+    pip uninstall lmdeploy -y
+    pip uninstall tensorflow -y
+    pip install optimum
+
+    # test with install
+    pip install .
+else
+    echo "Running case in release image, run case directly!"
+fi
+# remove torch_extensions folder to avoid ci hang.
+rm -rf ~/.cache/torch_extensions
+if [ $# -eq 0 ]; then
+    ci_command="pytest tests"
+else
+    ci_command="$@"
+fi
+echo "Running case with command: $ci_command"
+$ci_command

.dev_scripts/dockerci.sh

@@ -0,0 +1,96 @@
+#!/bin/bash
+MODELSCOPE_CACHE_DIR_IN_CONTAINER=/modelscope_cache
+CODE_DIR=$PWD
+CODE_DIR_IN_CONTAINER=/twinkle
+mkdir -p ~/.cache
+MODELSCOPE_CACHE=~/.cache
+IMAGE_NAME=modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/modelscope
+IMAGE_VERSION=ci_image
+MODELSCOPE_HOME_CACHE=~/.cache
+CI_TEST=True
+MODELSCOPE_SDK_DEBUG=True
+CI_COMMAND='bash .dev_scripts/ci_container_test.sh pytest tests'
+MODELSCOPE_SDK_DEBUG=True
+echo "$USER"
+gpus='0,1 2,3'
+cpu_sets='0-15 16-31'
+cpu_sets_arr=($cpu_sets)
+is_get_file_lock=false
+echo "ci command: $CI_COMMAND"
+PR_CHANGED_FILES="${PR_CHANGED_FILES:-}"
+echo "PR modified files: $PR_CHANGED_FILES"
+PR_CHANGED_FILES=${PR_CHANGED_FILES//[ ]/#}
+echo "PR_CHANGED_FILES: $PR_CHANGED_FILES"
+idx=0
+for gpu in $gpus
+do
+  exec {lock_fd}>"/tmp/gpu$gpu" || exit 1
+  flock -n "$lock_fd" || { echo "WARN: gpu $gpu is in use!" >&2; idx=$((idx+1)); continue; }
+  echo "get gpu lock $gpu"
+
+  CONTAINER_NAME="twinkle-ci-$idx"
+  let is_get_file_lock=true
+
+  # pull image if there are update
+  docker pull ${IMAGE_NAME}:${IMAGE_VERSION}
+  if [ "$MODELSCOPE_SDK_DEBUG" == "True" ]; then
+    echo 'debugging'
+    docker run --rm --name $CONTAINER_NAME --shm-size=16gb \
+              --cpuset-cpus=${cpu_sets_arr[$idx]} \
+              --gpus='"'"device=$gpu"'"' \
+              -v $CODE_DIR:$CODE_DIR_IN_CONTAINER \
+              -v $MODELSCOPE_CACHE:$MODELSCOPE_CACHE_DIR_IN_CONTAINER \
+              -v $MODELSCOPE_HOME_CACHE/$idx:/root \
+              -v /home/admin/pre-commit:/home/admin/pre-commit \
+              -e CI_TEST=True \
+              -e TEST_LEVEL=$TEST_LEVEL \
+              -e MODELSCOPE_CACHE=$MODELSCOPE_CACHE_DIR_IN_CONTAINER \
+              -e MODELSCOPE_DOMAIN=$MODELSCOPE_DOMAIN \
+              -e MODELSCOPE_SDK_DEBUG=True \
+              -e HUB_DATASET_ENDPOINT=$HUB_DATASET_ENDPOINT \
+              -e TEST_ACCESS_TOKEN_CITEST=$TEST_ACCESS_TOKEN_CITEST \
+              -e TEST_ACCESS_TOKEN_SDKDEV=$TEST_ACCESS_TOKEN_SDKDEV \
+              -e TEST_LEVEL=$TEST_LEVEL \
+              -e MODELSCOPE_ENVIRONMENT='ci' \
+              -e TEST_UPLOAD_MS_TOKEN=$TEST_UPLOAD_MS_TOKEN \
+              -e MODEL_TAG_URL=$MODEL_TAG_URL \
+              -e MODELSCOPE_API_TOKEN=$MODELSCOPE_API_TOKEN \
+	            -e PR_CHANGED_FILES=$PR_CHANGED_FILES \
+              --workdir=$CODE_DIR_IN_CONTAINER \
+              ${IMAGE_NAME}:${IMAGE_VERSION} \
+              $CI_COMMAND
+  else
+    docker run --rm --name $CONTAINER_NAME --shm-size=16gb \
+              --cpuset-cpus=${cpu_sets_arr[$idx]} \
+              --gpus='"'"device=$gpu"'"' \
+              -v $CODE_DIR:$CODE_DIR_IN_CONTAINER \
+              -v $MODELSCOPE_CACHE:$MODELSCOPE_CACHE_DIR_IN_CONTAINER \
+              -v $MODELSCOPE_HOME_CACHE/$idx:/root \
+              -v /home/admin/pre-commit:/home/admin/pre-commit \
+              -e CI_TEST=True \
+              -e TEST_LEVEL=$TEST_LEVEL \
+              -e MODELSCOPE_CACHE=$MODELSCOPE_CACHE_DIR_IN_CONTAINER \
+              -e MODELSCOPE_DOMAIN=$MODELSCOPE_DOMAIN \
+              -e HUB_DATASET_ENDPOINT=$HUB_DATASET_ENDPOINT \
+              -e TEST_ACCESS_TOKEN_CITEST=$TEST_ACCESS_TOKEN_CITEST \
+              -e TEST_ACCESS_TOKEN_SDKDEV=$TEST_ACCESS_TOKEN_SDKDEV \
+              -e TEST_LEVEL=$TEST_LEVEL \
+              -e MODELSCOPE_ENVIRONMENT='ci' \
+              -e TEST_UPLOAD_MS_TOKEN=$TEST_UPLOAD_MS_TOKEN \
+              -e MODEL_TAG_URL=$MODEL_TAG_URL \
+              -e MODELSCOPE_API_TOKEN=$MODELSCOPE_API_TOKEN \
+	            -e PR_CHANGED_FILES=$PR_CHANGED_FILES \
+              --workdir=$CODE_DIR_IN_CONTAINER \
+              ${IMAGE_NAME}:${IMAGE_VERSION} \
+              $CI_COMMAND
+  fi
+  if [ $? -ne 0 ]; then
+    echo "Running test case failed, please check the log!"
+    exit -1
+  fi
+  break
+done
+if [ "$is_get_file_lock" = false ] ; then
+    echo 'No free GPU!'
+    exit 1
+fi

.dev_scripts/dockerci_npu.sh

@@ -0,0 +1,57 @@
+#!/bin/bash
+MODELSCOPE_CACHE_DIR=/modelscope_cache
+CODE_DIR=$PWD
+MODELSCOPE_SDK_DEBUG=True
+echo "$USER"
+gpus='0,1 2,3'
+is_get_file_lock=false
+CI_COMMAND=${CI_COMMAND:-bash .dev_scripts/ci_container_test.sh pytest tests}
+echo "ci command: $CI_COMMAND"
+PR_CHANGED_FILES="${PR_CHANGED_FILES:-}"
+echo "PR modified files: $PR_CHANGED_FILES"
+PR_CHANGED_FILES=${PR_CHANGED_FILES//[ ]/#}
+echo "PR_CHANGED_FILES: $PR_CHANGED_FILES"
+idx=0
+for gpu in $gpus
+do
+  exec {lock_fd}>"/tmp/gpu$gpu" || exit 1
+  flock -n "$lock_fd" || { echo "WARN: gpu $gpu is in use!" >&2; idx=$((idx+1)); continue; }
+  echo "get gpu lock $gpu"
+
+  let is_get_file_lock=true
+
+  # 设置环境变量
+  export CI_TEST=True
+  export TEST_LEVEL=$TEST_LEVEL
+  export MODELSCOPE_CACHE=${MODELSCOPE_CACHE:-$MODELSCOPE_CACHE_DIR}
+  export MODELSCOPE_DOMAIN=$MODELSCOPE_DOMAIN
+  export HUB_DATASET_ENDPOINT=$HUB_DATASET_ENDPOINT
+  export TEST_ACCESS_TOKEN_CITEST=$TEST_ACCESS_TOKEN_CITEST
+  export TEST_ACCESS_TOKEN_SDKDEV=$TEST_ACCESS_TOKEN_SDKDEV
+  export MODELSCOPE_ENVIRONMENT='ci'
+  export TEST_UPLOAD_MS_TOKEN=$TEST_UPLOAD_MS_TOKEN
+  export MODEL_TAG_URL=$MODEL_TAG_URL
+  export MODELSCOPE_API_TOKEN=$MODELSCOPE_API_TOKEN
+  export PR_CHANGED_FILES=$PR_CHANGED_FILES
+  export CUDA_VISIBLE_DEVICES=$gpu
+
+  if [ "$MODELSCOPE_SDK_DEBUG" == "True" ]; then
+    export MODELSCOPE_SDK_DEBUG=True
+    echo 'debugging'
+  fi
+
+  # 切换到代码目录并执行命令
+  cd $CODE_DIR
+  eval $CI_COMMAND
+
+  if [ $? -ne 0 ]; then
+    echo "Running test case failed, please check the log!"
+    exit -1
+  fi
+  break
+done
+
+if [ "$is_get_file_lock" = false ] ; then
+    echo 'No free GPU!'
+    exit 1
+fi

.github/ISSUE_TEMPLATE/1-bug-report.yml

@@ -0,0 +1,49 @@
+name: "🐛 Bug Report"
+description: Create a bug report to help us improve twinkle
+labels: ["bug"]
+
+body:
+  - type: markdown
+    attributes:
+      value: |
+        Thank you for supporting twinkle and taking the time to submit this issue.
+        感谢你对 twinkle 的支持和抽出时间提交相关 issue。
+
+  - type: checkboxes
+    id: checklist
+    attributes:
+      label: Checklist / 检查清单
+      options:
+        - label: I have searched existing issues, and this is a new bug report. / 我已经搜索过现有的 issues，确认这是一个新的 bug report。
+          required: true
+
+
+  - type: textarea
+    id: bug-description
+    validations:
+      required: true
+    attributes:
+      label: Bug Description / Bug 描述
+      description: |
+        Please describe the issue you encountered. It's better to include error screenshots or stack trace information.
+        请详细描述你遇到的问题，最好包含报错截图或报错栈信息。
+
+
+  - type: textarea
+    id: reproduction-steps
+    validations:
+      required: true
+    attributes:
+      label: How to Reproduce / 如何复现
+      description: |
+        Please provide steps to reproduce the issue, including twinkle version, runtime environment, and detailed reproduction steps.
+        请提供复现问题的步骤，包括 twinkle 的版本、运行环境、详细的复现步骤等。
+
+
+  - type: textarea
+    id: additional-information
+    attributes:
+      label: Additional Information / 补充信息
+      description: |
+        Please provide any additional information here.
+        在这里补充其他相关信息。

.github/ISSUE_TEMPLATE/2-feature-request.yml

@@ -0,0 +1,37 @@
+name: "🚀 Feature Request"
+description: Submit a request for a new feature
+labels: ["enhancement"]
+
+body:
+  - type: markdown
+    attributes:
+      value: |
+        Thank you for supporting twinkle and taking the time to submit this issue.
+        感谢你对 twinkle 的支持和抽出时间提交相关 issue。
+
+  - type: checkboxes
+    id: checklist
+    attributes:
+      label: Checklist / 检查清单
+      options:
+        - label: I have searched existing issues, and this is a new feature request. / 我已经搜索过现有的 issues，确认这是一个新的 Feature Request。
+          required: true
+
+  - type: textarea
+    id: feature-request-description
+    validations:
+      required: true
+    attributes:
+      label: Feature Request Description / Feature Request 描述
+      description: |
+        Please provide a detailed description of the new feature you would like to see added.
+        请详细描述您希望添加的新功能特性。
+
+
+  - type: textarea
+    id: pull-request
+    attributes:
+      label: Pull Request / Pull Request 信息
+      description: |
+        Have you already submitted or plan to submit a Pull Request? Please share your plans.
+        你是否已经提交或即将提交 Pull Request？请说明你的计划。

.github/ISSUE_TEMPLATE/3-question-discussion.yml

@@ -0,0 +1,28 @@
+name: "🤔 Question & Discussion"
+description: Create an issue for questions and discussions
+labels: ["question"]
+
+body:
+  - type: markdown
+    attributes:
+      value: |
+        Thank you for supporting twinkle and taking the time to submit this issue.
+        感谢你对 twinkle 的支持和抽出时间提交相关 issue。
+
+  - type: checkboxes
+    id: checklist
+    attributes:
+      label: Checklist / 检查清单
+      options:
+        - label: I have searched existing issues, and this is a new question or discussion topic. / 我已经搜索过现有的 issues，确认这是一个新的问题与讨论。
+          required: true
+
+  - type: textarea
+    id: question-description
+    validations:
+      required: true
+    attributes:
+      label: Question Description / 问题描述
+      description: |
+        Please describe the question or topic you would like to discuss.
+        请描述你想要讨论的问题或话题。

.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1 @@
+blank_issues_enabled: false

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,13 @@
+# PR type
+- [ ] Bug Fix
+- [ ] New Feature
+- [ ] Document Updates
+- [ ] More Models or Datasets Support
+
+# PR information
+
+Write the detail information belongs to this PR.
+
+## Experiment results
+
+Paste your experiment result here(if needed).

.github/SECURITY.md

@@ -0,0 +1,3 @@
+# Reporting Security Issues
+
+Usually security issues of a deep learning project come from non-standard 3rd packages or continuous running services. If you are suffering from security issues from our project, please consider reporting to us. We appreciate your efforts to responsibly disclose your findings, and will make every effort to acknowledge your contributions.